David Heineman

DeepSeek R1 Distill Qwen 1.5B trained for 150 steps on a single env from https://arxiv.org/abs/2511.07317 (32 of 400 envs)